This is a predictor for the next 7 days that combines daily OHCL (open, high, low, and closing) data, twitter sentiment analysis and data from other top cryptocurrencies. "Red" means price falls in the following 7 days, "green" means increases in price and "grey" means "i don't known". The last colored dot is what we expect will happen (the chart shows previous predictions). The blue line represents the percent of variation according to the current price. The predictor uses deep learning techniques (combining state-of-the-art autoencoders with ensembles of LSTMs and MLPs) as well as feature selection methods. Accuracy of "greens" and "reds" are about 70% considering data from 2016 until now.
This is a predictor for the next hour that combines daily OHCL (open, high, low, and closing) data and data from other top cryptocurrencies. As the previous charts, "Red" means price falls in the following hour, "green" means increases in price and "grey" means "i don't known". The last colored dot is what we expect will happen (the chart shows previous predictions). The blue line represents the percent of variation according to the current price. The predictor uses deep learning techniques (combining state-of-the-art autoencoders with ensembles of LSTMs and MLPs) as well as feature selection methods. Accuracy of "greens" and "reds" are about 67% considering data from 2019 until now.
This third predictor evaluates the price tendency for the next 7 hours using machine learning techniques (mostly random forests) combined with sentiment analysis. Green/red dots represent the chance of increases/decreases in price. High distances between the current price (blue dot) and predictions mean higher chance. The last red/green is what we expect in the future.
This last predictor uses only sentiment analysis on twitter data to predict the price tendency for the next day. This exploitation gives an independet view of predictions consisting exclusively of twitter data.
This is a predictor for the next 6 hours that combines daily OHCL (open, high, low, and closing) data and NLP transformers on twitter data.
The predictors use machine learning (hourly predictions) and deep learning with autoencoders (daily predictions), following the ideas of [1] and [2]:
[1] Chen Z, Li C, Sun W (2020) Bitcoin price prediction using machine learning: an approach to sample dimension engineering. J Comput Appl Math 365:112395 https://www.sciencedirect.com/science/article/pii/S037704271930398X?casa_token=iuwHAcxOo3wAAAAA:2Tj7etHgoXdzqKfCQPGX0yMMmDZGQR0K0srxLRF3-_ZEPfr1B8OD2cFn4-iJ4nmdFkhEo5RRDkQ
[2] Zhang Z, Dai H, Garcia M (2021) Forecasting cryptocurrency price using convolutional neural networks with weighted and attentive memory channels. Expert Systems with Applications https://www.sciencedirect.com/science/article/abs/pii/S0957417421008046